Uniform bounded evaluation of the complete finite source reduction #
Two supplied unary counts bound the actual outer and candidate loops. Each call composes the sampler and selected affine program on the same original source. The two reductions then compute candidate XOR and strict majority, producing one bit on every input. The FP certificate includes parameter generation, binary advice encoding, and both bounded loops.
When the supplied counts equal the semantic counts and the growing sampler
guard holds, the program computes sourceReductionExtractor exactly. The
evaluator does not generate unrestricted exponential counts from their
widths; choosing bounded counts for an eventual family is a separate layer.
The construction follows Chattopadhyay--Liao, Extractors for Sum of Two
Sources, Lemma 5.4 and the proof of Theorem 2:
https://arxiv.org/abs/2110.12652.
The runtime preserves the supplied number of outer rows.
Every parameter choice and source word produces one verdict bit.
The majority of no outer rows follows the false tie convention.
Empty candidate rows all have false XOR, hence the verdict is false.
Exact binary enumeration computes every semantic XOR coordinate in its original order.
Exact supplied counts and the sampler guard identify the actual finite construction.
Supplied polynomial-time unary counts make both actual enumeration loops polynomial-time.
Decode the source word and all three supplied unary parameters.
Malformed paired words also produce exactly one bit.
The canonical paired evaluator computes the exact semantic Boolean output.
One total paired program includes all parameter, call, XOR, and majority computations.